Skip to main content

Selecting a datasource

Every conversation asks its questions in one place: a datasource, and within it usually a catalog and a schema. You can start from the Catalog, which shows the three levels as a tree — datasource → catalog → schema — and change the place at any time from the chip by the composer.

This page covers what each level means, how to open a schema in chat, how to see and change where a conversation asks, and what to do when you don't see what you expect.

The three levels​

Datasource A connection your administrator configured: PostgreSQL, BigQuery, Snowflake, Starburst, Databricks, SAP HANA, Oracle, SQL Server, Redshift, Cloudera, Salesforce, etc. You can only see datasources you've been granted access to via RBAC.

Catalog The top-level grouping inside a datasource. The exact meaning is database-specific:

  • BigQuery — a project.
  • Snowflake / Databricks — a database / catalog.
  • PostgreSQL — a database.
  • SAP HANA — a database + module combination (see SAP module integration for details).

Schema The collection of tables your queries actually run against. When a conversation has a schema, Qry works from that schema's tables.

Steps​

1. Open the Catalog​

Click Catalog in the navigation (the database icon, under Data). The Catalog is a workbench: the tree of datasources on the left, the detail of whatever you select on the right.

The Catalog: datasources in the tree, nothing selected yet

The tree lists every datasource your account can browse, each with its engine on the right. Filter tables at the top finds a table by name across the datasources you have opened.

2. Open a datasource​

Click a datasource. The tree expands to its catalogs, and the right pane shows how many tables are cached for it and when its schema was last synced, with Refresh schema to read it again.

A datasource opened: its catalogs in the tree, its schema status on the right

3. Drill into a catalog and schema​

Expand a catalog to see its schemas, and click the schema you want to work with.

A schema selected: its tables listed on the right

The right pane summarises the schema — tables, views, how many are profiled, the datasource — and lists its tables. Click a table to see its columns and preview a sample of its data without leaving the Catalog.

4. Open in chat​

Click Open in chat at the top right. Qry starts a new conversation that asks in this schema.

Open in chat on a table does the same and starts the question for you: the prompt reads Tell me about the table with the table pinned as a chip, so Qry reads that table's columns. Finish the sentence or send it as it is. (Pinning works as with @ — see Pinning data and skills.)

Where a conversation asks​

A chip by the composer always shows where the conversation's questions go:

  • On a computer or tablet, in the composer's footer next to the model: the datasource, then catalog.schema — for example Sales DWH · analytics.orders.
  • On a phone, one chip above the composer with the schema, followed by the model only when it is not your usual one.

Click the chip to change it. On a computer it opens Where Qry asks; on a phone, This conversation, which also lets you change the model. Under A specific place, choose a Data source, a Catalog and a Schema, and click Done (or close the sheet). See it in the catalog opens the same place in the Catalog.

Where Qry asks, on A specific place: Data source, Catalog and Schema, with See it in the catalog and Done.

What Qry gets depends on how far you narrow it:

You chooseThe chip readsWhat Qry does
A schemaDatasource · catalog.schemaWorks from that schema's tables.
A catalog, All its schemasDatasource · catalog · all its schemasGets the tables of every schema in the catalog (a shortened list when there are many).
No catalogDatasource · everything you can queryLooks through every catalog and schema you are allowed to query and finds its way. It works, but it is slower and can pick the wrong schema — choosing a catalog makes answers faster and more precise.

The change applies from your next question, also in the middle of a conversation. When you reopen a conversation, it comes back to where its last answer asked — the domain or the schema — even if you have been working somewhere else since.

Asking in a domain​

Your administrators can gather the tables of a data area under the name people use — Hotel bookings rather than ventas_dw.reservas — as a business domain. When you may ask in at least one, Where Qry asks (on a phone, This conversation) starts with a choice: A domain or A specific place. Under A domain, pick one: under it you see where its data lives and what it holds, and the chip shows its name.

Where Qry asks, on A domain: Clientes IXEN chosen, its sources (BigQuery + Trino) and what it holds; the chip reads Clientes IXEN.

A question asked in a domain gives Qry only the domain's tables — those you may query — and a note saying what the domain holds, so answers are faster and more precise than asking over a whole catalog. You see only the domains where you can reach at least one schema; a domain never gives you access to a table you could not query anyway.

You can also mention a domain in the question with @ — domains are offered first, above tables — and that question asks in it; the conversation stays there.

Typing @clie: the domains first ("ask this question in it"), each with where its data lives, then the tables you can query.

Domains appear in Search or ask Qry (⌘K) too, and open the Catalog by domain.

Search or ask Qry for "clientes": tables, then the domains, notebooks and dashboards that match.

A specific place goes back to data source, catalog and schema. Choosing a place any other way — the schema explorer, a workspace's defaults, pinning a table from another catalog — also leaves the domain. If a domain stops being available to you (it was deleted, or its audience no longer includes you), your next question is not sent: Qry says so and leaves the domain, so you can choose where to ask.

tip

A conversation's history refers to the tables it has used. If you move it to another schema and ask for "the same again", answers can be confusing when table or column names don't line up — for an unrelated schema, a new conversation is usually clearer.

Common issues​

The datasource I expected isn't in the tree. You don't have an RBAC grant for it. Contact your administrator. See Connecting databases for the admin side.

The datasource is there but the catalog tree is empty. You have datasource-level access but no catalog-level grants. Some tenants configure RBAC at the catalog or schema level — your admin can extend or refine the grant.

A table is in the schema but a query against it errors with "permission denied". Table-level grants and ABAC tag policies operate independently of schema-level access. Your admin can tell you which rule is hiding the row or column.

A new table doesn't appear. The schema is cached so that browsing never wakes the engine. New tables are picked up within the hour, or the first time a query names one; Refresh schema on the datasource reads it again right away.

See also​